buildwellai-model-v2 / scripts /prepare_dataset.py
Choukrijer's picture
Upload BuildwellAI Qwen3-14B project
499c907 verified
#!/usr/bin/env python3
"""
BuildwellAI Model V2 - Complete Dataset Preparation Script
Combines all data sources:
1. BSI Flex 8670 Building Safety Competence Benchmark (CSV)
2. UK Building Control Competency Benchmark 2025 (CSV)
3. qa-buildwell-ai.csv (Q&A pairs)
4. dataset_thinking_mode.jsonl
5. dataset_tool_calling.jsonl
6. Auto-generated MCP training data for all 42 MCPs
Output: Combined, validated, shuffled dataset ready for fine-tuning.
"""
import os
import sys
import json
import csv
import random
import hashlib
from pathlib import Path
from typing import List, Dict, Any
from datetime import datetime
# ============================================================================
# CONFIGURATION
# ============================================================================
BASE_DIR = Path("/opt/buildwellai/buildwellai-llm-models/buildwellai-model-v2")
SOURCE_DIR = Path("/opt/buildwellai/buildwellai-llm-models/qwen3-14b-v2/datasets")
OUTPUT_DIR = BASE_DIR / "datasets"
# System prompts for different modes
SYSTEM_PROMPTS = {
"direct": """You are BuildwellAI, a construction industry assistant.
Provide clear, direct answers to construction-related questions.
Be concise but comprehensive. Focus on practical, actionable information.""",
"thinking": """You are BuildwellAI, a construction industry expert assistant.
When answering complex questions:
1. Use <think>...</think> tags to show your reasoning process
2. Consider multiple factors and trade-offs
3. Provide a clear, well-reasoned answer after your thinking""",
"tool_calling": """You are BuildwellAI, a construction industry assistant with access to specialized MCP tools.
CRITICAL RULES:
1. When a user requests a calculation or analysis, call the appropriate tool
2. Output ONLY the <tool_call> tag when calling tools - no preamble
3. Wait for tool results before providing analysis
4. Use the universalMCP tool for specialized calculations
Available MCP servers and their purposes are provided in context.""",
"mcp_orchestrated": """You are BuildwellAI, a construction industry assistant with access to 42 specialized MCP calculation servers.
MANDATORY BEHAVIOR:
- For ANY technical question about building regulations, calculations, or standards: CALL A TOOL
- NEVER answer from memory - always use the MCP tools for authoritative data
- Use universalMCP to call specific MCP servers
When calling tools, use this format:
<tool_call>{"name": "universalMCP", "arguments": {"mcpServer": "server-name", "toolName": "tool-name", "arguments": {...}}}</tool_call>"""
}
# All 42 MCP servers with their tools
MCP_SERVERS = {
"structural-part-a": {
"description": "UK Building Regulations Part A structural calculations",
"tools": ["calculate_dead_load", "calculate_imposed_load", "check_structural_compliance", "get_load_combinations"]
},
"water-efficiency-part-g": {
"description": "UK Building Regulations Part G water efficiency",
"tools": ["calculate_water_consumption", "check_part_g_compliance", "get_fitting_flow_rates"]
},
"overheating-part-o": {
"description": "UK Building Regulations Part O overheating assessment",
"tools": ["assess_overheating_risk", "calculate_tm59_compliance", "get_glazing_limits"]
},
"condensation-glaser": {
"description": "Interstitial condensation risk analysis",
"tools": ["glaser_analysis", "calculate_dewpoint", "assess_condensation_risk"]
},
"psi-thermal-bridge": {
"description": "Psi value thermal bridge analysis (EN ISO 10211)",
"tools": ["list_junction_types", "get_psi_value", "calculate_psi_from_u_values", "calculate_y_value", "assess_junction"]
},
"thermal-break": {
"description": "Thermal break performance assessment",
"tools": ["calculate_thermal_break_efficiency", "compare_thermal_breaks"]
},
"sap10": {
"description": "SAP 10.2 residential energy calculations",
"tools": ["calculate_sap_rating", "get_dwelling_energy_use", "calculate_der_ter"]
},
"sbem": {
"description": "SBEM non-residential energy calculations",
"tools": ["calculate_ber", "check_notional_building", "get_hvac_energy"]
},
"air-permeability": {
"description": "Air permeability test data processing",
"tools": ["calculate_air_permeability", "check_part_l_compliance", "estimate_infiltration_heat_loss"]
},
"drainage-swmm": {
"description": "Drainage runoff calculations (SWMM methodology)",
"tools": ["calculate_runoff", "design_pipe_network", "simulate_storm_event"]
},
"suds": {
"description": "SuDS flow and volume calculations (CIRIA C753)",
"tools": ["design_suds", "calculate_attenuation", "check_discharge_rate"]
},
"hydraulic-drainage": {
"description": "Hydraulic drainage network analysis",
"tools": ["analyze_drainage_network", "calculate_pipe_capacity", "check_gradients"]
},
"embodied-carbon": {
"description": "Embodied carbon assessment (ICE Database 2023)",
"tools": ["calculate_embodied_carbon", "get_material_carbon_factors", "compare_materials"]
},
"smoke-ventilation": {
"description": "Smoke ventilation design (BS 7346, BS EN 12101)",
"tools": ["design_smoke_shaft", "calculate_aov_size", "check_smoke_clearance"]
},
"evacuation": {
"description": "Evacuation time modelling (PD 7974-6)",
"tools": ["calculate_evacuation_time", "model_egress_flow", "assess_phased_evacuation"]
},
"disproportionate-collapse": {
"description": "Disproportionate collapse assessment (EN 1991-1-7)",
"tools": ["check_robustness_class", "calculate_tie_forces", "assess_key_elements"]
},
"sound-insulation": {
"description": "Sound insulation calculations (Approved Document E)",
"tools": ["calculate_sound_reduction", "check_part_e_compliance", "estimate_flanking"]
},
"fire-safety": {
"description": "Fire safety assessment (Approved Document B)",
"tools": ["calculate_travel_distance", "check_compartmentation", "assess_means_of_escape", "get_fire_resistance_requirements"]
},
"flood-risk": {
"description": "Flood risk assessment (NPPF, FEH)",
"tools": ["calculate_flood_risk", "determine_flood_zone", "assess_sequential_test"]
},
"biodiversity-net-gain": {
"description": "Biodiversity net gain calculations (Environment Act 2021)",
"tools": ["calculate_bng_units", "assess_habitat_condition", "check_10_percent_gain"]
},
"daylight-factor": {
"description": "Daylight factor calculations (BS 8206-2)",
"tools": ["calculate_daylight_factor", "check_room_depth", "assess_no_sky_line"]
},
"adf-modelling": {
"description": "Average daylight factor modelling (BS EN 17037)",
"tools": ["calculate_adf", "check_target_illuminance", "model_daylight_distribution"]
},
"sunlight-overshadowing": {
"description": "Sunlight/overshadowing analysis (BRE 209)",
"tools": ["calculate_sunlight_hours", "assess_overshadowing", "check_bre_guidelines"]
},
"cwct-facade": {
"description": "CWCT facade performance testing",
"tools": ["check_weather_tightness", "calculate_structural_capacity", "assess_thermal_movement"]
},
"dynamic-thermal": {
"description": "Dynamic thermal simulation",
"tools": ["simulate_annual_energy", "calculate_peak_loads", "optimize_hvac_sizing"]
},
"breeam": {
"description": "BREEAM sustainability assessment",
"tools": ["calculate_breeam_credits", "get_category_requirements", "assess_rating_path"]
},
"well": {
"description": "WELL Building Standard assessment",
"tools": ["check_well_features", "calculate_optimization_credits", "assess_preconditions"]
},
"passivhaus": {
"description": "Passivhaus/PHPP calculations",
"tools": ["calculate_heating_demand", "check_airtightness", "assess_phpp_balance", "get_standard_requirements"]
},
"contaminated-land": {
"description": "Contaminated land assessment (CLEA)",
"tools": ["calculate_gac", "assess_risk_pathway", "determine_remediation"]
},
"cfd-fire-smoke": {
"description": "CFD fire/smoke simulation",
"tools": ["simulate_smoke_spread", "calculate_visibility", "assess_tenability"]
},
"wufi-hygrothermal": {
"description": "WUFI hygrothermal analysis",
"tools": ["simulate_moisture_transport", "assess_mould_risk", "check_interstitial_condensation"]
},
"lca": {
"description": "Life Cycle Assessment (EN 15978, RICS 2023)",
"tools": ["calculate_whole_life_carbon", "get_lca_benchmarks", "compare_design_options"]
},
"hydraulic-transport": {
"description": "Hydraulic transport analysis",
"tools": ["calculate_slurry_flow", "design_pneumatic_conveying", "assess_pipe_wear"]
},
"daylighting": {
"description": "Daylighting design and analysis",
"tools": ["optimize_glazing_ratio", "calculate_solar_gain", "model_light_distribution"]
},
"energy-modelling": {
"description": "Building energy modelling",
"tools": ["calculate_annual_energy", "optimize_fabric", "compare_hvac_options"]
},
"indoor-air-quality": {
"description": "Indoor air quality assessment",
"tools": ["calculate_ventilation_rate", "assess_co2_levels", "check_pollutant_concentrations"]
},
"pedestrian-comfort": {
"description": "Pedestrian comfort analysis (Lawson criteria)",
"tools": ["assess_wind_comfort", "check_lawson_criteria", "identify_mitigation"]
},
"wind-microclimate": {
"description": "Wind microclimate assessment",
"tools": ["model_wind_patterns", "calculate_amplification", "assess_safety"]
},
"vibration": {
"description": "Ground-borne vibration assessment (BS 6472-1)",
"tools": ["calculate_vdv", "assess_human_perception", "check_building_damage_threshold"]
},
"swept-path": {
"description": "Vehicle swept path analysis",
"tools": ["calculate_swept_path", "check_vehicle_clearance", "optimize_turning_radius"]
},
"carbon-offset": {
"description": "Carbon offset calculations",
"tools": ["calculate_offset_requirement", "estimate_offset_cost", "compare_offset_options"]
},
"ventilation-part-f": {
"description": "UK Building Regulations Part F ventilation compliance",
"tools": ["calculate_ventilation_rates", "check_part_f_compliance", "design_mev_system"]
}
}
# ============================================================================
# DATA GENERATION FUNCTIONS
# ============================================================================
def generate_mcp_training_examples() -> List[Dict]:
"""Generate training examples for all 42 MCP servers with realistic prompts."""
examples = []
# MCP-specific realistic prompts - what professionals actually ask
MCP_PROMPTS = {
"psi-thermal-bridge": [
"What are the SAP Appendix K junction types?",
"List all the thermal bridge junction codes",
"What's the PSI value for junction E5?",
"I need the default PSI value for a ground floor junction",
"Calculate the Y-value for my dwelling with these junctions",
"What PSI value should I use for a steel lintel?",
"Show me all the external wall junction types",
"What's the temperature factor for junction E1?",
"I have a balcony thermal bridge - what junction code is that?",
"What's the difference between E5 and E19 junctions?",
"Calculate total heat loss from thermal bridges",
"What PSI values apply to party wall junctions?",
"I need to assess thermal bridging at the roof-wall junction",
"What's the enhanced construction detail PSI for E10?",
"List roof junction types R1 to R11",
],
"sap10": [
"Calculate the SAP rating for a 3-bed semi-detached house",
"What's the target emission rate for my new dwelling?",
"I need to run a SAP 10.2 calculation",
"What's the dwelling emission rate for this property?",
"Calculate primary energy consumption for the dwelling",
"Will this house achieve EPC band B?",
"What's the fabric energy efficiency for my design?",
"I need to check Part L compliance for a new build",
"Calculate space heating demand",
"What's the carbon emission rate?",
"Run SAP assessment for a 2-bed flat",
"What hot water energy consumption should I expect?",
"Calculate the notional dwelling values",
"What improvements would get this to EPC A?",
"Assess the dwelling against Part L 2021",
],
"breeam": [
"What BREEAM credits can I achieve for Ene 01?",
"Calculate BREEAM score for this office building",
"What are the mandatory credits for BREEAM Excellent?",
"How many Wat 01 credits for 40% water reduction?",
"What's required for BREEAM Outstanding?",
"List all credits in the Energy category",
"What exemplary credits are available?",
"Calculate minimum standards for BREEAM Very Good",
"What Hea 02 credits for daylight compliance?",
"I need to hit BREEAM Excellent - what's the pathway?",
"What Mat 01 credits for responsible sourcing?",
"Calculate Land Use and Ecology credits",
"What innovation credits can we claim?",
"Assess transport accessibility credits",
"What Pol 03 credits for flood risk?",
],
"fire-safety": [
"What's the maximum travel distance for an open plan office?",
"Calculate escape route capacity for 500 occupants",
"What fire resistance is required for a 12-storey building?",
"Is a sprinkler system required for this building?",
"What compartment sizes apply to this retail unit?",
"Calculate the number of escape stairs needed",
"What's the minimum door width for 200 people?",
"Does this care home need a stay-put strategy?",
"What fire stopping is required at service penetrations?",
"Calculate fire engineering parameters for BS 9999",
"What protected stairway width do I need?",
"Is a firefighting shaft required?",
"What external wall fire spread requirements apply?",
"Calculate occupancy for this restaurant",
"What emergency lighting is required?",
],
"passivhaus": [
"What's the heating demand limit for Passivhaus Classic?",
"Calculate airtightness requirement for certification",
"What's the primary energy demand limit?",
"Does this design meet Passivhaus Plus criteria?",
"What window U-values are needed for Passivhaus?",
"Calculate the thermal bridge free design requirement",
"What MVHR efficiency is required?",
"Check if overheating criteria are met",
"What's the specific space heating demand?",
"Calculate renewable energy generation requirements",
"What's the maximum heating load?",
"Does this meet PHI Low Energy Building standard?",
"What treated floor area should I use?",
"Calculate the frequency of overheating",
"What Passivhaus windows are suitable for this climate?",
],
"embodied-carbon": [
"Calculate embodied carbon for this concrete frame",
"What's the carbon factor for structural steel?",
"Compare embodied carbon of CLT vs concrete",
"What's the A1-A3 carbon for this building?",
"Calculate whole life carbon including B and C stages",
"What carbon savings from using recycled aggregate?",
"What's the embodied carbon benchmark for offices?",
"Calculate carbon for foundation concrete",
"Compare brick vs timber frame carbon impact",
"What's the GWP for mineral wool insulation?",
"Calculate carbon sequestration for timber structure",
"What RICS benchmarks apply to residential?",
"What's the upfront carbon for this design?",
"Calculate carbon intensity per square metre",
"What are the LETI 2030 targets?",
],
"condensation-glaser": [
"Run a Glaser condensation risk assessment",
"Is there interstitial condensation risk in this wall?",
"Check dewpoint temperature through the construction",
"Will moisture accumulate in the insulation layer?",
"Calculate vapour pressure through the wall buildup",
"Is a vapour barrier needed for this roof?",
"What's the condensation risk for a cold bridged detail?",
"Check internal surface condensation risk",
"Calculate critical surface humidity",
"What's the monthly moisture balance?",
"Is there mould growth risk on internal surfaces?",
"Check timber frame wall for condensation",
"What ventilated cavity width prevents condensation?",
"Calculate interstitial humidity levels",
"Assess warm roof vs cold roof condensation risk",
],
"overheating-part-o": [
"Check Part O compliance for this bedroom",
"Calculate overheating risk using TM59",
"Does the living room pass the overheating criteria?",
"What glazing ratio is allowed facing south-west?",
"Calculate hours above 26°C threshold",
"Is cross-ventilation sufficient for Part O?",
"What shading is needed to pass overheating?",
"Run TM52 assessment for this office",
"Calculate adaptive comfort temperature",
"What percentage of occupied hours exceed threshold?",
"Check if night-time ventilation is adequate",
"What g-value glazing is needed?",
"Calculate bedroom overheating for 32°C night",
"Does this flat need mechanical cooling?",
"Assess overheating for a top floor flat",
],
"structural-part-a": [
"Calculate dead and imposed loads for this floor",
"What load combination factors apply?",
"Check beam capacity for this loading",
"Calculate foundation bearing pressure",
"What's the maximum span for 47x200 joists at 400 centres?",
"Check deflection limits for this beam",
"Calculate wind load on the gable wall",
"What steel section is needed for 6m span?",
"Check Part A compliance for this structure",
"Calculate load transfer through the structure",
"What's the shear capacity of this connection?",
"Check lateral stability requirements",
"Calculate roof load including snow",
"What padstone size is needed under this beam?",
"Check wall stability for this height",
],
"lca": [
"Run whole life carbon assessment",
"Calculate lifecycle carbon stages A-C",
"What's the operational energy over 60 years?",
"Compare design options for lifecycle impact",
"What module D benefits apply?",
"Calculate end of life carbon",
"What's the use stage carbon for this building?",
"Apply EN 15978 assessment methodology",
"Calculate replacement cycles for cladding",
"What's the reference study period?",
"Compare lifecycle cost and carbon",
"What RICS 2023 methodology should I use?",
"Calculate maintenance phase carbon",
"What system boundaries apply?",
"Assess refurbishment vs new build LCA",
],
"sound-insulation": [
"What Rw is needed between these flats?",
"Calculate airborne sound insulation",
"Does this wall meet Part E requirements?",
"What's the DnT,w requirement for this dwelling?",
"Calculate impact sound insulation for floor",
"What L'nT,w should the floor achieve?",
"Check flanking transmission paths",
"What's the sound reduction index for this partition?",
"Calculate pre-completion testing requirements",
"Does this separating wall meet Robust Details?",
"What acoustic performance for party floor?",
"Calculate Ctr correction for traffic noise",
"What internal wall sound insulation is needed?",
"Check dwelling entrance door acoustic rating",
"What ceiling treatment improves impact sound?",
],
"daylight-factor": [
"Calculate average daylight factor for this room",
"Does the room meet 2% daylight factor?",
"What glazing ratio for adequate daylight?",
"Check no-sky line position in the room",
"Calculate daylight uniformity ratio",
"What room depth is acceptable for daylight?",
"Check BS EN 17037 target illuminance",
"Calculate daylight factor at back of room",
"What external obstruction angle affects daylight?",
"Does this classroom meet BB101 daylight?",
"Calculate minimum daylight to kitchen",
"What ADF is required for living room?",
"Check daylight distribution across floor plate",
"What glazing transmittance for 2% DF?",
"Calculate supplementary electric lighting need",
],
"smoke-ventilation": [
"Design smoke shaft for this residential building",
"Calculate AOV size for this lobby",
"What smoke clearance rate is required?",
"Size the natural smoke vent for this stair",
"What mechanical extract rate for car park?",
"Calculate smoke reservoir depth",
"What inlet air is needed for balanced design?",
"Check BS EN 12101-6 design parameters",
"Size smoke extract fan capacity",
"What free area is needed for smoke ventilation?",
"Calculate pressure differential for lobby",
"Design depressurisation system for firefighting stair",
"What smoke control for single escape stair?",
"Calculate make-up air velocity",
"Size emergency ventilation shaft",
],
"evacuation": [
"Calculate total evacuation time for this building",
"What's the flow rate through this exit?",
"Model phased evacuation for 20-storey building",
"Calculate pre-movement time for office",
"What's the merging flow at the final exit?",
"Check stair capacity for simultaneous evacuation",
"Calculate required safe egress time",
"What's the available safe egress time?",
"Model evacuation for care home residents",
"Calculate queue formation at exit doors",
"What egress time for mobility impaired?",
"Check refuge area sizing",
"Calculate evacuation with phased alarm",
"What's the walking speed on stairs?",
"Model evacuation with one exit blocked",
],
"water-efficiency-part-g": [
"Calculate water consumption per person per day",
"Does this dwelling meet 125 litres/person/day?",
"What's the optional water target of 110 lpppd?",
"Calculate water use for dual flush WC",
"What flow rate for Part G compliant taps?",
"Check fittings against water calculator",
"What shower flow rate meets the target?",
"Calculate dwelling water consumption",
"What water fittings achieve 110 target?",
"Run Part G water efficiency calculation",
"What bath size for water compliance?",
"Calculate greywater recycling benefit",
"What rainwater harvesting credits apply?",
"Check washing machine water consumption",
"Calculate potable water demand",
],
"ventilation-part-f": [
"Calculate whole dwelling ventilation rate",
"What extract rates for kitchen and bathroom?",
"Does this ventilation strategy meet Part F?",
"Size the MVHR unit for this dwelling",
"What background ventilator size is needed?",
"Calculate minimum purge ventilation",
"What continuous extract for System 3?",
"Check trickle ventilator equivalent area",
"What's the air supply rate per person?",
"Size extract fan for utility room",
"Calculate Part F for open plan kitchen",
"What crossflow ventilation opening size?",
"Check passive stack ventilation sizing",
"What boost extract rate for cooker hood?",
"Calculate MEV system airflow rates",
],
"air-permeability": [
"What air permeability target for Part L?",
"Calculate heat loss from infiltration",
"What's the design air permeability?",
"Process air test results at 50Pa",
"What air tightness for Passivhaus?",
"Calculate infiltration heat loss coefficient",
"What backstop air permeability applies?",
"Check pulse test results",
"What's the equivalent leakage area?",
"Calculate annual infiltration heat loss",
"What improvements reduce air leakage?",
"Process blower door test data",
"What ACH50 is required?",
"Calculate shelter factor effect",
"What air barrier strategy is recommended?",
],
"biodiversity-net-gain": [
"Calculate baseline biodiversity units",
"What's the 10% BNG requirement?",
"Calculate habitat condition score",
"What strategic significance multiplier applies?",
"Run biodiversity metric 4.0 calculation",
"What hedgerow units are on site?",
"Calculate off-site offset requirement",
"What habitat creation achieves 10% gain?",
"Check distinctiveness score for grassland",
"Calculate river units for watercourse",
"What temporal multiplier for new habitat?",
"Assess difficulty of habitat creation",
"What's the habitat trading rules?",
"Calculate net change in biodiversity",
"What management plan is required?",
],
"flood-risk": [
"What flood zone is this site in?",
"Is this development appropriate in Flood Zone 3?",
"Calculate finished floor level for flood risk",
"What sequential test applies?",
"Run exception test for vulnerable development",
"What flood resilience measures are needed?",
"Calculate 1% AEP flood level",
"What climate change allowance applies?",
"Check EA flood map data",
"What's the surface water flood risk?",
"Calculate freeboard requirement",
"What FRA scope for major development?",
"Is flood compensation storage required?",
"What groundwater flood risk exists?",
"Calculate flood storage loss and compensation",
],
"suds": [
"Design attenuation tank for 1 in 100 year storm",
"Calculate greenfield runoff rate",
"What SuDS features achieve 4 pillars?",
"Size permeable paving for car park",
"Calculate swale dimensions",
"What detention basin volume is needed?",
"Design rain garden for treatment",
"Calculate long term storage requirement",
"What interception storage is required?",
"Size filter drain for access road",
"What SuDS treatment train is needed?",
"Calculate half drain time",
"Design bioretention system",
"What infiltration rate should I use?",
"Calculate discharge rate to watercourse",
],
"well": [
"What WELL features meet Air concept?",
"Calculate WELL certification pathway",
"What preconditions are mandatory?",
"How many optimizations for WELL Gold?",
"Check WELL v2 Light requirements",
"What Nourishment features apply to office?",
"Calculate WELL points for this project",
"What's required for WELL Platinum?",
"Check Movement concept requirements",
"What Thermal Comfort features needed?",
"Calculate Sound concept compliance",
"What Materials preconditions apply?",
"Check Mind concept features",
"What Community features for workplace?",
"Calculate Innovation points available",
],
"disproportionate-collapse": [
"What robustness class is this building?",
"Calculate horizontal tie forces",
"Check notional element removal",
"What key element design is required?",
"Calculate vertical tie requirements",
"Is the building Class 2B?",
"What peripheral ties are needed?",
"Check alternative load path",
"Calculate internal tie forces",
"What corner column ties required?",
"Design for accidental loading",
"Check Eurocode 1991-1-7 requirements",
"What is the tie force for 7.5m grid?",
"Calculate column removal scenario",
"What enhanced design for Class 3?",
],
"dynamic-thermal": [
"Run annual energy simulation",
"Calculate peak heating load",
"What cooling load for this office?",
"Model fabric heat loss",
"Calculate solar gains through glazing",
"What HVAC sizing from simulation?",
"Run IES model for energy",
"Calculate thermal mass effect",
"What's the annual heating demand?",
"Model natural ventilation strategy",
"Calculate comfort hours",
"What plant capacity is needed?",
"Run summer overheating simulation",
"Calculate energy use intensity",
"Model mixed-mode ventilation",
],
"sbem": [
"Calculate BER for this office building",
"What's the target emission rate for SBEM?",
"Run Part L2A compliance check",
"Calculate notional building values",
"What's the primary energy from SBEM?",
"Check modular building approach",
"Calculate heating and cooling loads",
"What EPC rating will this achieve?",
"Run SBEM for shell and core",
"What improvement factor is needed?",
"Calculate actual building emission rate",
"Check HVAC system efficiencies",
"What lighting efficacy for compliance?",
"Calculate renewable contribution",
"Run Part L2B assessment for extension",
],
"thermal-break": [
"Calculate thermal break efficiency for steel beam",
"What thermal break is needed for a balcony connection?",
"Compare Schock Isokorb vs generic thermal break",
"What temperature factor does this thermal break achieve?",
"Calculate heat loss reduction from thermal break",
"What PSI value improvement from thermal break?",
"Size thermal break for cantilevered balcony",
"What thermal break for steel to concrete connection?",
"Calculate fRsi at thermal break location",
"What structural capacity for Isokorb KXT?",
"Compare thermal break performance ratings",
"What thermal break for parapet connection?",
"Calculate linear thermal transmittance with break",
"What thermal break depth is required?",
"Assess condensation risk at thermal break",
],
"drainage-swmm": [
"Calculate peak runoff for 1:100 year storm",
"Model drainage network capacity",
"What pipe size for this catchment?",
"Simulate storm event in SWMM",
"Calculate time of concentration",
"Design pipe network for development",
"What's the runoff coefficient for this site?",
"Model combined sewer overflow",
"Calculate hydraulic grade line",
"What storage is needed for attenuation?",
"Simulate 30-year storm event",
"Check pipe capacity under climate change",
"Calculate critical storm duration",
"Model surcharge in network",
"Design outfall to watercourse",
],
"hydraulic-drainage": [
"Analyze drainage network hydraulics",
"Check pipe gradient for self-cleansing",
"Calculate pipe capacity for foul drainage",
"What flow velocity in this pipe?",
"Design gravity drainage system",
"Check hydraulic capacity of existing drain",
"Calculate Manning's equation for pipe flow",
"What pipe diameter for 2 l/s flow?",
"Assess gradient for 100mm pipe",
"Calculate roughness coefficient",
"Design pumped drainage system",
"Check ventilation for drainage system",
"Calculate design flow from dwellings",
"What fall is needed for 150mm drain?",
"Assess surcharge capacity",
],
"adf-modelling": [
"Calculate average daylight factor for office",
"What ADF for living room compliance?",
"Model daylight distribution across floorplate",
"Does the room meet BS EN 17037 targets?",
"Calculate uniformity ratio for daylight",
"What glazing area for 2% ADF?",
"Model combined sky and sun exposure",
"Calculate median daylight illuminance",
"What target ADF for kitchen?",
"Assess daylight with external obstruction",
"Calculate supplementary lighting need",
"Model daylight for open plan office",
"What room depth for adequate daylight?",
"Calculate ADF at work plane height",
"Assess glare probability",
],
"sunlight-overshadowing": [
"Calculate annual probable sunlight hours",
"Does the garden get BRE 209 compliant sunlight?",
"Assess overshadowing from proposed building",
"What's the 25% APSH for south-facing window?",
"Calculate winter sunlight (Dec-Feb)",
"Model shadow path diagram",
"Assess sunlight to neighbouring property",
"Calculate sun on ground at spring equinox",
"What overshadowing impact on amenity space?",
"Check 2-hour sun on March 21",
"Model transient overshadowing",
"Calculate loss of sunlight to existing building",
"What APSH reduction is acceptable?",
"Assess sunlight access to playground",
"Model sun hours for rooftop amenity",
],
"cwct-facade": [
"Check CWCT weather tightness classification",
"Calculate facade wind load capacity",
"What mullion deflection under wind?",
"Assess thermal movement in cladding",
"Design glass size for wind pressure",
"Check structural silicone joint capacity",
"Calculate bracket spacing for facade",
"What tolerance for facade installation?",
"Assess impact resistance requirements",
"Design facade for CWCT test sequence A",
"Calculate air permeability for curtain wall",
"What water penetration resistance?",
"Check facade for fire spread requirements",
"Calculate thermal bridge at transom",
"Assess dynamic wind load on facade",
],
"contaminated-land": [
"Calculate generic assessment criteria",
"What soil guideline values apply?",
"Assess risk pathway for residential use",
"Determine CLEA screening level",
"What remediation target for lead?",
"Calculate human health risk assessment",
"Check groundwater protection requirements",
"What asbestos screening level?",
"Assess vapour intrusion pathway",
"Calculate averaging area for sampling",
"What clean cover depth required?",
"Determine Category 4 screening level",
"Assess source-pathway-receptor linkage",
"What remediation verification needed?",
"Calculate bioaccessibility factor",
],
"cfd-fire-smoke": [
"Model smoke spread in atrium",
"Calculate visibility during fire scenario",
"Simulate tenability conditions",
"What smoke layer height over time?",
"Model smoke extraction effectiveness",
"Calculate temperature at head height",
"Simulate fire growth rate impact",
"What make-up air velocity needed?",
"Model smoke flow through opening",
"Calculate carbon monoxide concentration",
"Assess safe escape conditions",
"Simulate sprinkler activation effect",
"What smoke density at 2m height?",
"Model smoke movement in car park",
"Calculate buoyancy-driven flow",
],
"wufi-hygrothermal": [
"Simulate moisture transport in wall",
"Assess mould growth risk with WUFI",
"Check interstitial condensation over year",
"What relative humidity in insulation?",
"Model timber moisture content cycle",
"Calculate drying potential of wall",
"Simulate vapour barrier performance",
"What moisture accumulation over 5 years?",
"Assess warm side vapour control",
"Model rain-driven moisture ingress",
"Calculate hygrothermal performance",
"What critical RH threshold for mould?",
"Simulate renovation wall performance",
"Assess breathable membrane effectiveness",
"Model cold bridge moisture risk",
],
"hydraulic-transport": [
"Calculate slurry flow in pipe",
"Design pneumatic conveying system",
"What settling velocity for particles?",
"Assess pipe wear from abrasive material",
"Calculate critical deposition velocity",
"Design vacuum transport system",
"What pressure drop in slurry pipe?",
"Assess solids concentration limit",
"Calculate hindered settling velocity",
"Design dense phase conveying",
"What pipe size for sand slurry?",
"Assess erosion rate in bend",
"Calculate transport velocity for coal",
"Design plug flow conveying",
"What air velocity for pneumatic transport?",
],
"daylighting": [
"Optimize glazing ratio for daylight",
"Calculate solar gain coefficient",
"Design light shelf for deeper daylight",
"What window size for adequate light?",
"Model reflectance for daylight distribution",
"Calculate daylight penetration depth",
"Design skylight for top lighting",
"What glazing VLT for daylight?",
"Model combined side and top lighting",
"Calculate view factor to sky",
"Design clerestory for even light",
"What external shading for solar control?",
"Model light pipe for windowless room",
"Calculate useful daylight illuminance",
"Design atrium for daylight distribution",
],
"energy-modelling": [
"Calculate annual heating energy demand",
"Model building energy consumption",
"What fabric improvements reduce energy?",
"Compare HVAC system energy use",
"Calculate energy use intensity",
"Model renewable energy contribution",
"What baseload energy for building?",
"Calculate peak demand for sizing",
"Model occupancy impact on energy",
"What U-value changes affect energy most?",
"Calculate operational carbon from energy",
"Model heat pump COP impact",
"What lighting energy for office?",
"Calculate hot water energy demand",
"Model plug load energy consumption",
],
"indoor-air-quality": [
"Calculate ventilation rate for IAQ",
"What CO2 level indicates good IAQ?",
"Assess PM2.5 filtration requirement",
"Check pollutant concentration limits",
"Calculate air changes for fresh air",
"What VOC level is acceptable?",
"Design filtration for urban location",
"Calculate formaldehyde from materials",
"What outside air rate per person?",
"Assess radon mitigation requirement",
"Calculate pollutant source strength",
"What HEPA filtration for healthcare?",
"Design demand-controlled ventilation",
"Calculate NOx ingress from traffic",
"What air cleaning for school?",
],
"pedestrian-comfort": [
"Assess wind comfort at ground level",
"Check Lawson sitting comfort criteria",
"What wind mitigation for entrance?",
"Calculate pedestrian safety threshold",
"Design wind screen for seating area",
"Check comfort for outdoor dining",
"What wind conditions for thoroughfare?",
"Assess corner acceleration effect",
"Calculate downdraft from tall building",
"Design canopy for wind protection",
"Check comfort for balcony space",
"What wind criteria for playground?",
"Assess venturi effect between buildings",
"Calculate gust equivalent mean",
"Design landscaping for wind mitigation",
],
"wind-microclimate": [
"Model wind patterns around building",
"Calculate amplification factor for wind",
"Assess pedestrian safety for high rise",
"What wind acceleration between towers?",
"Model corner effect wind speed",
"Calculate wind comfort probability",
"Assess channelling effect in street",
"What wind conditions at podium level?",
"Model downwash from building",
"Calculate reference wind speed",
"Assess wind for rooftop amenity",
"What direction dominates wind?",
"Model mitigation with porous screen",
"Calculate wind comfort hours",
"Assess existing conditions baseline",
],
"vibration": [
"Calculate vibration dose value",
"Assess human perception of vibration",
"Check building damage from vibration",
"What VDV limit for residential?",
"Calculate vibration from railway",
"Design vibration isolation for plant",
"What PPV threshold for listed building?",
"Assess groundborne vibration path",
"Calculate vibration transfer to structure",
"Design floating floor for vibration",
"What vibration criteria for hospital?",
"Assess piling vibration impact",
"Calculate vibration from road traffic",
"Design resilient mounting for equipment",
"What vibration isolation efficiency?",
],
"swept-path": [
"Calculate swept path for fire tender",
"Design turning circle for refuse vehicle",
"What clearance for articulated lorry?",
"Check swept path for car park",
"Design entrance for large delivery vehicle",
"Calculate tracking width on bend",
"What corner radius for bus?",
"Assess reversing manoeuvre space",
"Design loading bay access",
"Calculate tail swing clearance",
"What geometry for coach drop-off?",
"Assess fire appliance access route",
"Calculate swept path for ambulance",
"Design turning head for cul-de-sac",
"What dimensions for HGV access?",
],
"carbon-offset": [
"Calculate carbon offset requirement",
"What offset cost per tonne CO2?",
"Compare offset options for development",
"Design carbon reduction before offset",
"What residual carbon needs offsetting?",
"Calculate offset fund contribution",
"Check local authority offset rate",
"What carbon payback period?",
"Calculate net zero carbon pathway",
"Design to minimize offset requirement",
"What quality criteria for offsets?",
"Calculate annual offset requirement",
"Design renewable energy for offset credit",
"What additionality requirement?",
"Calculate lifetime carbon for offset",
],
}
# Generic prompts for MCPs without specific ones
generic_prompts = [
"Calculate {metric} for {context}",
"What's the {metric} for {context}?",
"I need to determine {metric} for {context}",
"Check {topic} compliance for {context}",
"Assess {topic} for {context}",
"Run {topic} analysis on {context}",
"What are the requirements for {context}?",
"Help me with {topic} calculation for {context}",
"Can you analyze {topic} for {context}?",
"I need a {topic} assessment for {context}",
"What {topic} data do you have for {context}?",
"Perform {topic} check on {context}",
"Calculate the {metric} requirements",
"What standards apply to {context}?",
"Evaluate {context} for {topic}",
]
# Rich context examples for different domains
contexts = {
"structural": [
"a 6m steel beam", "a residential floor slab", "a load-bearing wall", "a roof truss",
"a concrete column", "a timber joist", "a steel portal frame", "a reinforced concrete beam",
"a masonry wall", "a cantilever balcony", "a basement retaining wall", "a floor diaphragm",
"a steel connection", "a precast panel", "a composite slab", "a foundation pad"
],
"thermal": [
"a cavity wall", "a flat roof", "a ground floor slab", "a timber frame wall",
"a solid wall with external insulation", "a warm roof construction", "a cold roof",
"a party wall", "a window reveal", "a door threshold", "a floor-wall junction",
"an insulated concrete form wall", "a curtain wall system", "a roof-wall junction",
"a steel lintel detail", "a balcony thermal break"
],
"energy": [
"a 3-bedroom dwelling", "a 500m2 office", "a school building", "a retail unit",
"a 2-bed flat", "a 4-bed detached house", "a care home", "a student accommodation",
"a warehouse", "a restaurant", "a leisure centre", "a hospital ward",
"a hotel", "a data centre", "a laboratory", "a community centre"
],
"fire": [
"an 8-storey residential building", "an open-plan office", "a shopping centre", "a hotel",
"a care home", "a school", "a hospital", "a warehouse", "a mixed-use development",
"a basement car park", "a high-rise apartment block", "a student accommodation",
"a cinema", "a nightclub", "a sports hall", "an underground station"
],
"sustainability": [
"a new office development", "a residential scheme", "a mixed-use building",
"a school refurbishment", "a hospital extension", "a retail park", "a logistics hub",
"a university campus", "an affordable housing project", "a commercial fit-out",
"a heritage conversion", "a net-zero housing development", "a science park"
],
"drainage": [
"a 2-hectare site", "a commercial car park", "a housing development",
"a school playground", "an industrial estate", "a retail car park",
"a highway drainage scheme", "a sports field", "a green roof system",
"a permeable paving area", "a flood attenuation basin", "a soakaway"
],
"daylight": [
"a deep plan office", "a residential living room", "a classroom",
"a hospital ward", "an atrium space", "a basement flat", "a north-facing room",
"an open plan kitchen-diner", "a bedroom", "a retail showroom"
],
"acoustics": [
"a party wall between dwellings", "a floor separating flats", "an internal partition",
"a classroom wall", "a hotel room", "a recording studio", "a conference room",
"a residential development near a railway", "an office near a busy road"
]
}
# Regulations for compliance queries
regulations = [
"Building Regulations Part A", "Building Regulations Part B", "Building Regulations Part E",
"Building Regulations Part F", "Building Regulations Part G", "Building Regulations Part L",
"Building Regulations Part M", "Building Regulations Part O", "Building Regulations Part S",
"BS 9999", "BS 5950", "Eurocode 2", "Eurocode 3", "Approved Document B",
"BREEAM requirements", "Passivhaus standards", "WELL certification", "Part L 2021",
"SAP 10.2", "SBEM compliance", "TM59 overheating criteria", "BRE 209 guidelines"
]
# Metrics for calculation queries
metrics = {
"structural": ["dead load", "imposed load", "ultimate load", "serviceability deflection", "shear capacity", "bending moment"],
"thermal": ["U-value", "PSI value", "Y-value", "thermal resistance", "condensation risk", "temperature factor"],
"energy": ["SAP rating", "EPC band", "primary energy", "CO2 emissions", "heating demand", "cooling load"],
"fire": ["travel distance", "fire resistance period", "smoke clearance time", "evacuation time", "compartment size"],
"sustainability": ["BREEAM score", "embodied carbon", "operational carbon", "whole life carbon", "biodiversity units"],
"drainage": ["runoff rate", "attenuation volume", "discharge rate", "flood risk level", "pipe capacity"],
}
# Topics for assessment queries
topics = {
"structural": ["structural capacity", "stability", "robustness", "load path", "connection strength"],
"thermal": ["thermal performance", "condensation risk", "thermal bridging", "heat loss", "moisture transfer"],
"energy": ["energy efficiency", "carbon performance", "compliance pathway", "fabric performance", "ventilation strategy"],
"fire": ["fire safety", "means of escape", "compartmentation", "smoke control", "structural fire resistance"],
"sustainability": ["sustainability performance", "environmental impact", "carbon footprint", "biodiversity impact"],
"drainage": ["drainage capacity", "flood risk", "surface water management", "SuDS performance"],
}
def get_domain(mcp_name: str) -> str:
"""Determine domain from MCP name."""
if any(x in mcp_name for x in ["structural", "collapse", "vibration"]):
return "structural"
elif any(x in mcp_name for x in ["thermal", "psi", "condensation", "u-value", "wufi", "hygro"]):
return "thermal"
elif any(x in mcp_name for x in ["sap", "sbem", "energy", "passivhaus", "dynamic", "air-perm"]):
return "energy"
elif any(x in mcp_name for x in ["fire", "smoke", "evacuation", "cfd"]):
return "fire"
elif any(x in mcp_name for x in ["breeam", "well", "carbon", "lca", "biodiversity", "embodied"]):
return "sustainability"
elif any(x in mcp_name for x in ["drainage", "suds", "flood", "hydraulic", "water"]):
return "drainage"
elif any(x in mcp_name for x in ["daylight", "adf", "sunlight"]):
return "daylight"
elif any(x in mcp_name for x in ["sound", "acoustic"]):
return "acoustics"
else:
return random.choice(["structural", "thermal", "energy"])
# Generate ~100+ examples per MCP for robust training
for mcp_name, mcp_info in MCP_SERVERS.items():
description = mcp_info["description"]
tools = mcp_info["tools"]
domain = get_domain(mcp_name)
# Get domain-specific data
domain_contexts = contexts.get(domain, contexts["structural"])
domain_metrics = metrics.get(domain, metrics["structural"])
domain_topics = topics.get(domain, topics["structural"])
# Get MCP-specific realistic prompts if available
specific_prompts = MCP_PROMPTS.get(mcp_name, [])
# 1. Generate examples from specific realistic prompts (highest quality)
for prompt in specific_prompts:
# Match prompt to most appropriate tool
best_tool = tools[0] # Default to first tool
prompt_lower = prompt.lower()
for tool in tools:
tool_words = tool.replace("_", " ").lower().split()
if any(word in prompt_lower for word in tool_words):
best_tool = tool
break
tool_call = {
"name": "universalMCP",
"arguments": {
"mcpServer": mcp_name,
"toolName": best_tool,
"arguments": {}
}
}
example = {
"messages": [
{"role": "system", "content": SYSTEM_PROMPTS["mcp_orchestrated"]},
{"role": "user", "content": prompt},
{"role": "assistant", "content": f'<tool_call>{json.dumps(tool_call)}</tool_call>'}
]
}
examples.append(example)
# Add variations of the same prompt
variations = [
f"I need help with: {prompt}",
f"Quick question - {prompt.lower()}",
f"For my project, {prompt.lower()}",
f"Can you help me? {prompt}",
prompt.replace("?", ""), # Statement form
]
for var in variations[:3]: # Add 3 variations per specific prompt
var_example = {
"messages": [
{"role": "system", "content": SYSTEM_PROMPTS["mcp_orchestrated"]},
{"role": "user", "content": var},
{"role": "assistant", "content": f'<tool_call>{json.dumps(tool_call)}</tool_call>'}
]
}
examples.append(var_example)
# 2. Generate additional examples using templates for each tool
for tool in tools:
tool_name_readable = tool.replace("_", " ")
# Generate 8 template-based examples per tool
for _ in range(8):
context = random.choice(domain_contexts)
metric = random.choice(domain_metrics)
topic = random.choice(domain_topics)
regulation = random.choice(regulations)
template = random.choice(generic_prompts)
user_query = template.format(
topic=topic,
context=context,
metric=metric,
regulation=regulation
)
tool_call = {
"name": "universalMCP",
"arguments": {
"mcpServer": mcp_name,
"toolName": tool,
"arguments": {}
}
}
example = {
"messages": [
{"role": "system", "content": SYSTEM_PROMPTS["mcp_orchestrated"]},
{"role": "user", "content": user_query},
{"role": "assistant", "content": f'<tool_call>{json.dumps(tool_call)}</tool_call>'}
]
}
examples.append(example)
# 3. Add examples with tool results (important for complete conversations)
for i, tool in enumerate(tools):
tool_name_readable = tool.replace("_", " ")
context = random.choice(domain_contexts)
# Use specific prompt if available
if specific_prompts and i < len(specific_prompts):
user_query = specific_prompts[i]
else:
user_query = f"Can you {tool_name_readable} for {context}?"
tool_call = {
"name": "universalMCP",
"arguments": {
"mcpServer": mcp_name,
"toolName": tool,
"arguments": {}
}
}
# Realistic tool result
result_text = f"Analysis completed successfully. The {tool_name_readable} for {context} shows the following results: [detailed calculation output would appear here]. The values meet the applicable standards and regulations."
example_with_result = {
"messages": [
{"role": "system", "content": SYSTEM_PROMPTS["mcp_orchestrated"]},
{"role": "user", "content": user_query},
{"role": "assistant", "content": f'<tool_call>{json.dumps(tool_call)}</tool_call>'},
{"role": "tool", "content": json.dumps({"success": True, "result": result_text})},
{"role": "assistant", "content": f"Based on the {tool_name_readable} analysis, here are the key findings:\n\n1. The calculation has been completed for {context}\n2. Results comply with the relevant {domain} standards\n3. The {mcp_name.replace('-', ' ')} assessment confirms acceptable performance\n\nWould you like me to provide more details on any specific aspect?"}
]
}
examples.append(example_with_result)
return examples
def convert_csv_to_jsonl(csv_path: Path, mode: str = "direct") -> List[Dict]:
"""Convert CSV Q&A pairs to JSONL training format."""
examples = []
# Read file and remove NUL characters
with open(csv_path, 'rb') as f:
content = f.read().replace(b'\x00', b'').decode('utf-8-sig', errors='ignore')
import io
reader = csv.DictReader(io.StringIO(content))
for row in reader:
# Handle different CSV formats
question = row.get('Question') or row.get('question') or row.get('Q', '')
answer = row.get('Answer') or row.get('answer') or row.get('A') or row.get('Correct Answer', '')
if not question or not answer:
continue
question = question.strip()
answer = answer.strip()
if len(question) < 10 or len(answer) < 10:
continue
# Choose system prompt based on complexity
if len(answer) > 200 or any(x in question.lower() for x in ['explain', 'how', 'why', 'compare']):
system_prompt = SYSTEM_PROMPTS["thinking"]
# Add thinking wrapper for complex answers
answer = f"<think>\nThis question about {question[:50]}... requires careful analysis.\nLet me consider the key factors.\n</think>\n\n{answer}"
else:
system_prompt = SYSTEM_PROMPTS["direct"]
example = {
"messages": [
{"role": "system", "content": system_prompt},
{"role": "user", "content": question},
{"role": "assistant", "content": answer}
]
}
examples.append(example)
return examples
def load_jsonl(path: Path) -> List[Dict]:
"""Load existing JSONL file."""
examples = []
with open(path, 'r', encoding='utf-8') as f:
for line in f:
line = line.strip()
if line:
try:
examples.append(json.loads(line))
except json.JSONDecodeError:
continue
return examples
def validate_example(example: Dict) -> bool:
"""Validate a training example has correct format."""
if "messages" not in example:
return False
messages = example["messages"]
if not isinstance(messages, list) or len(messages) < 2:
return False
# Check for valid roles
valid_roles = {"system", "user", "assistant", "tool"}
for msg in messages:
if not isinstance(msg, dict):
return False
if "role" not in msg or "content" not in msg:
return False
if msg["role"] not in valid_roles:
return False
# Ensure at least one user message
has_user = any(m["role"] == "user" for m in messages)
has_assistant = any(m["role"] == "assistant" for m in messages)
return has_user and has_assistant
def deduplicate_examples(examples: List[Dict]) -> List[Dict]:
"""Remove duplicate examples based on content hash."""
seen = set()
unique = []
for ex in examples:
# Create hash of user content
user_content = "".join(
m["content"] for m in ex["messages"]
if m["role"] == "user"
)
content_hash = hashlib.md5(user_content.encode()).hexdigest()
if content_hash not in seen:
seen.add(content_hash)
unique.append(ex)
return unique
def balance_dataset(examples: List[Dict], max_per_type: int = 15000) -> List[Dict]:
"""Balance dataset to prevent overfitting on any single type."""
categorized = {
"direct": [],
"thinking": [],
"tool_calling": [],
"mcp": []
}
for ex in examples:
system = ex["messages"][0]["content"] if ex["messages"][0]["role"] == "system" else ""
assistant = ex["messages"][-1]["content"] if ex["messages"][-1]["role"] == "assistant" else ""
if "universalMCP" in assistant or "mcpServer" in assistant:
categorized["mcp"].append(ex)
elif "<tool_call>" in assistant:
categorized["tool_calling"].append(ex)
elif "<think>" in assistant:
categorized["thinking"].append(ex)
else:
categorized["direct"].append(ex)
balanced = []
for cat, items in categorized.items():
if len(items) > max_per_type:
balanced.extend(random.sample(items, max_per_type))
else:
balanced.extend(items)
return balanced
# ============================================================================
# MAIN PROCESSING
# ============================================================================
def main():
print("=" * 70)
print("BuildwellAI Model V2 - Dataset Preparation")
print("=" * 70)
print(f"Output directory: {OUTPUT_DIR}")
print()
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
all_examples = []
stats = {}
# 1. Load existing JSONL datasets
print("Loading existing JSONL datasets...")
thinking_path = SOURCE_DIR / "dataset_thinking_mode.jsonl"
if thinking_path.exists():
thinking = load_jsonl(thinking_path)
all_examples.extend(thinking)
stats["thinking_mode"] = len(thinking)
print(f" - dataset_thinking_mode.jsonl: {len(thinking):,} examples")
tool_calling_path = SOURCE_DIR / "dataset_tool_calling.jsonl"
if tool_calling_path.exists():
tools = load_jsonl(tool_calling_path)
all_examples.extend(tools)
stats["tool_calling"] = len(tools)
print(f" - dataset_tool_calling.jsonl: {len(tools):,} examples")
# 2. Convert CSV files
print("\nConverting CSV files...")
csv_files = [
("qa-buildwell-ai.csv", "qa_pairs"),
("UK Building Control Competency Benchmark 2025 - Benchmark.csv", "uk_benchmark"),
("BSI Flex 8670 Building Safety Competence Benchmark - generate a benchmark for all these questiosn.csv", "bsi_benchmark"),
]
for csv_name, stat_key in csv_files:
csv_path = SOURCE_DIR / csv_name
if csv_path.exists():
converted = convert_csv_to_jsonl(csv_path)
all_examples.extend(converted)
stats[stat_key] = len(converted)
print(f" - {csv_name}: {len(converted):,} examples")
else:
print(f" - {csv_name}: NOT FOUND")
# 3. Generate MCP training data
print("\nGenerating MCP training data for 42 servers...")
mcp_examples = generate_mcp_training_examples()
all_examples.extend(mcp_examples)
stats["mcp_generated"] = len(mcp_examples)
print(f" - Generated: {len(mcp_examples):,} MCP examples")
# 4. Validate all examples
print("\nValidating examples...")
valid_examples = [ex for ex in all_examples if validate_example(ex)]
invalid_count = len(all_examples) - len(valid_examples)
print(f" - Valid: {len(valid_examples):,}")
print(f" - Invalid (removed): {invalid_count:,}")
# 5. Deduplicate
print("\nDeduplicating...")
unique_examples = deduplicate_examples(valid_examples)
dupes_removed = len(valid_examples) - len(unique_examples)
print(f" - Unique: {len(unique_examples):,}")
print(f" - Duplicates removed: {dupes_removed:,}")
# 6. Balance dataset
print("\nBalancing dataset...")
balanced_examples = balance_dataset(unique_examples, max_per_type=15000)
print(f" - Balanced: {len(balanced_examples):,}")
# 7. Shuffle
print("\nShuffling...")
random.seed(42)
random.shuffle(balanced_examples)
# 8. Split into train/val
val_size = int(len(balanced_examples) * 0.05) # 5% validation
train_examples = balanced_examples[val_size:]
val_examples = balanced_examples[:val_size]
print(f"\nFinal split:")
print(f" - Training: {len(train_examples):,}")
print(f" - Validation: {len(val_examples):,}")
# 9. Save datasets
print("\nSaving datasets...")
train_path = OUTPUT_DIR / "train.jsonl"
with open(train_path, 'w', encoding='utf-8') as f:
for ex in train_examples:
f.write(json.dumps(ex, ensure_ascii=False) + '\n')
print(f" - {train_path}")
val_path = OUTPUT_DIR / "validation.jsonl"
with open(val_path, 'w', encoding='utf-8') as f:
for ex in val_examples:
f.write(json.dumps(ex, ensure_ascii=False) + '\n')
print(f" - {val_path}")
# Save combined for reference
combined_path = OUTPUT_DIR / "combined.jsonl"
with open(combined_path, 'w', encoding='utf-8') as f:
for ex in balanced_examples:
f.write(json.dumps(ex, ensure_ascii=False) + '\n')
print(f" - {combined_path}")
# 10. Save statistics
stats_path = OUTPUT_DIR / "dataset_stats.json"
final_stats = {
"generated_at": datetime.now().isoformat(),
"source_counts": stats,
"total_before_processing": len(all_examples),
"valid_examples": len(valid_examples),
"unique_examples": len(unique_examples),
"balanced_examples": len(balanced_examples),
"train_examples": len(train_examples),
"val_examples": len(val_examples),
"mcp_servers_covered": len(MCP_SERVERS),
}
with open(stats_path, 'w') as f:
json.dump(final_stats, f, indent=2)
print(f" - {stats_path}")
print("\n" + "=" * 70)
print("DATASET PREPARATION COMPLETE")
print("=" * 70)
print(f"\nTotal training examples: {len(train_examples):,}")
print(f"Ready for fine-tuning!")
if __name__ == "__main__":
main()